When Should You Use Demonstration Recordings, Videos, and Other Performances?

Building a Piano Piece · Season Two: From One Piece to Long-Term Progress · Article Nine

Listening before learning can quickly reveal approximate tempo, sound and proportion. It can also turn another performer’s phrasing, dynamics and timing into defaults before the learner has read the score. Video exposes movement, but can invite copying of visible shape without understanding the problem the movement solves.

The debate is often reduced to “imitation is harmful” or “listening to masters naturally improves playing”. Both ignore stage and purpose. An external performance is most valuable as a source of comparable sound, movement clues and possibilities. Whether it supports independence depends on the question asked before watching, the return to score and personal performance afterwards, and comparison beyond one model.

Research shows short-term effects, not a final interpretation

A 2026 study randomly assigned 54 undergraduate and graduate piano majors to modelling, structural-understanding and narrative-imagery approaches while practising the opening of Chopin’s E-flat major Nocturne. The modelling group reported heightened awareness of phrasing, rubato, accent and balance cues, while the approaches produced different results and experiences. One sixty-minute session with one excerpt cannot establish that long-term imitation forms mature style, or that modelling is superior to structure and imagery.

An action-observation experiment asked 21 musically untrained participants to watch an expert perform a short keyboard melody and imitate it, comparing this with watching keys light in sequence. Observation of the player assisted some measures of correct keys, force similarity and note duration. It was a short-sequence novice study. It does not establish that every wrist angle in a professional pianist’s video should be copied.

Listening to one’s own performance is another form of externalised feedback. A study of 25 advanced piano students compared Disklavier playback, adjustable playback and no playback. Students hearing playback improved some self-ratings, while professional external ratings did not show significant group effects. The authors cautiously concluded that playback influenced self-observation. Hearing a problem once is not proof that the next performance has objectively improved.

Together, these findings support a bounded judgement: models and playback can change attention, expectation and self-evaluation. Learning still needs independent execution, delayed retrieval, varied context and audible results.

Four materials, four uses

The score supplies the musical text, edition information and possible composer or editor markings. It does not automatically produce one sound.

Audio concentrates attention on timing, timbre, voices, pedal and large proportion while removing visible movement.

Performance video adds information about hands, arms, body, breathing, relocation and stage organisation. Camera angle, editing and bodily differences limit inference.

Personal audio or video is not a model but an external observation position. Microphone placement, automatic gain and phone processing alter sound; not every dynamic relation reproduces the room.

Ask first whether the problem needs text, a sound possibility, movement information or self-feedback. More information does not make video best for every purpose.

Stage one: establish a minimal independent reading

When independent reading is the goal, spend ten to fifteen minutes mapping structure, imagining sound and making a controlled first trial before hearing a complete performance. The model then lands on existing questions instead of constructing the learner’s first map.

An exception arises when style or notation is deeply unfamiliar, or an ensemble relation cannot be understood from one part. A short early excerpt may establish the type of sound. Return to the score and distinguish textual information from the performer’s decisions.

Prepare three questions before listening: What are the principal proportions? Which texture cannot yet be imagined? Which tempo or pedal marking needs an audible result? Undirected listening can create familiarity while producing few practice decisions.

Stage two: compare instead of obeying

Choose at least two reliable, high-quality versions with perceptibly different approaches. Compare one dimension at a time: opening tempo, repeats, local timing, melody–inner-voice balance, pedal clarity or location of climax.

Use three columns:

“I like A” becomes an audible relationship. The table also acknowledges that recorded performances are not answers in a right–wrong test. Edition, instrument, recording period and tradition may explain differences. Ornament, repeat and textual questions require reliable editions and scholarship rather than majority voting among recordings.

Stage three: extract one cue, then remove the model

Dependence develops when a learner listens and copies phrase by phrase, unable to proceed without playback. Use a short listen–describe–attempt–remove cycle.

Select ten to twenty seconds and hear it twice. Away from the keyboard, describe the relationship: “The highest melody note is not the loudest; the left hand lightens before the harmony changes; time expands at the cadence without losing pulse.” Turn off the recording, locate structural support or limits in the score, and design a personal trial. Record it several minutes later without replaying the model.

For movement, state the problem solved. Where do the eyes move before a leap? When is a repeated chord released? How does the arm transport the hand? Do not copy wrist height or finger shape as an object. Hand, span, bench, key response and tempo differ. Visible form must survive testing against personal sound, control and comfort.

Stage four: separate your recording from the model

Give each personal recording one question. First listen without the score and record large direction; listen again with the score to locate events; only then compare with an external version. Simultaneous judgement of pitch, timbre, rhythm, pedal, movement and resemblance destroys priority.

Change one variable and record again. Later or next day, cold-test before hearing the model. If tempo, phrase and movement vanish without the recording, the current performance remains cue-dependent.

Personal playback is not an absolute judge. In the study of 25 advanced students, improved self-ratings did not align with significant external-rating changes. The correct conclusion is “I have acquired another observation position”, not “playback guarantees that attempt two is better”.

When to listen early, later or near completion

Earlier listening may help when style or notation is unfamiliar, ensemble context is missing, the texture cannot be internally heard, historical instruments are being compared, or a teacher has assigned modelling.

Independent reading should usually come first when sight-reading and decoding are the goal, the form is legible, or the learner readily copies tempo unconsciously.

Mid-process listening is useful after a precise problem appears in voicing, pedal, timing or movement strategy.

Near-completion listening tests large proportion and acquired habits. Multiple versions are especially useful then because the learner possesses enough structure to resist automatic copying.

Three common misuses

First, treating view counts, fame or competition results as interpretive authority. None replaces source, history and audible reasoning.

Second, using slow-motion video as movement diagnosis. It reveals sequence and route but not internal force, causes of pain or coordination. Persistent discomfort requires appropriate teaching and health assessment, not imitation of another body.

Third, counting listening as practice without producing an output. Active listening leaves a description, comparison, score mark or new performance test. Background exposure can create familiarity but is not equivalent to analysis or motor learning.

A model-use record

Record six items: source and version; question before listening; relationship observed; support or limitation in the score; personal experiment; cold test after model removal. An empty final field means demonstration has not yet become independent capacity.

Recordings and videos should neither be forbidden nor allowed to replace judgement. Their strongest place is in the middle of a cycle: the learner forms a question, external performance enlarges the possibilities, and the work returns to text, body and personal sound.

A version is worth studying not because it tells you how the piece must finally be played, but because it reveals a question you had not asked and requires a testable reason for your own answer.

Primary sources and further reading

• Comparison of modelling, structural understanding and narrative imagery in independent piano practice
• Action-observation training and novice keyboard skill
• Disklavier audio-playback feedback for advanced piano students
• Listening and score-reading in recognition of musical structure

Continue reading: Building a Piano Piece series page

If you would like to bring these ideas about listening, understanding, and practice to the keyboard, you might try ScoreFlow, an app I developed to make score reading and daily practice flow more naturally together.